Application of a method for correcting an observed regression between change and initial value for the bias caused by random errors in the initial value

Abstract

When studying the correlation between the change of a variable during treatment and the value before treatment, random errors (errors of measurement as well as intraindividual biological variation) may yield seriously biased results. In the present work a method to adjust for this bias is presented. The method was applied to data from blood coagulation factor and plasma protein analysis in chronic alcoholics before and after 1 week of abstinence. It was found that many of the significant correlations were lost when the data were adjusted.